Quantile classification
A way to group data into classes for a map so that each class holds the same number of features, whatever the range of values inside it.
Before a map can colour areas by a value, the values have to be grouped into classes, one colour per class. Quantile classification sets the class breaks so that every class holds the same number of features. With five classes, each one holds a fifth of the data: the lowest fifth gets the lightest colour, the highest fifth the darkest.
The method is easiest to see next to equal interval classification, which splits the range of values into steps of the same size. Take ten areas with these values: 2, 3, 4, 5, 6, 7, 9, 12, 25 and 60.
| Class | Quantile: range | Quantile: areas | Equal interval: range | Equal interval: areas |
|---|---|---|---|---|
| 1 | 2 to 3 | 2 | 2 to 13.6 | 8 |
| 2 | 4 to 5 | 2 | 13.6 to 25.2 | 1 |
| 3 | 6 to 7 | 2 | 25.2 to 36.8 | 0 |
| 4 | 9 to 12 | 2 | 36.8 to 48.4 | 0 |
| 5 | 25 to 60 | 2 | 48.4 to 60 | 1 |
With equal intervals, eight of the ten areas share one colour and two classes stay empty. The map shows one outlier and little else. With quantiles, every colour is used and the differences between the lower values become visible. That is why quantiles work well for ranking: they answer which areas are in the top fifth.
The cost is that class widths stop meaning anything. In the example, 25 and 60 share a class, while 7 and 9 are in different ones. A quantile map can make small differences look large and large ones look small, so the legend has to show the actual ranges. For data with a few extreme values, heads and tails can be the better choice. In GOAT, quantile classification is one of six methods in attribute-based styling, with seven classes by default.